Create-knowledge-find-decode/acquisition/search.py

"""关键词搜索：query → content_id 列表（翻页 + 去重 + 截断）。

接口（实测 2026-06-26，host = settings.crawler_base_url）：
    小红书 POST /crawler/xiao_hong_shu/keyword
      body{keyword, content_type:"图文"|"视频", sort_type:"综合", publish_time:"", cursor:""}
      → {code:0, data:{has_more, next_cursor, data:[{id:<content_id>, note_card:{...}}]}}
    抖音   POST /crawler/dou_yin/keyword
      body{keyword, content_type:"视频", sort_type:"综合排序", publish_time:"不限",
           cursor:"0", account_id:"771431222"}   # ← 缺 account_id 会 code 10000
      → {code:0, data:{has_more, next_cursor, data:[{aweme_id:<content_id>, ...}]}}

各平台请求参数 / id 字段不同，集中在 PLATFORM_SEARCH 表里。
搜索只给预览；全文/全图/视频仍需 crawler.fetch_post_detail 逐 id 取详情（detail 只需 content_id）。
parse_search_response 纯函数可离线测；search_keyword 负责 HTTP + 翻页。
"""
from __future__ import annotations

import re
from typing import Any, Optional
from urllib.parse import urljoin

import httpx

from core.config import Settings
from acquisition.crawler import RateLimiter

# 每平台的搜索参数差异（path / 条目 id 字段 / 排序 / 发布时间 / 起始 cursor / account_id）
PLATFORM_SEARCH = {
    "xiaohongshu": {
        "path": "/crawler/xiao_hong_shu/keyword", "id_key": "id",
        "sort_type": "综合", "publish_time": "", "cursor0": "", "account_id": None,
    },
    "douyin": {
        "path": "/crawler/dou_yin/keyword", "id_key": "aweme_id",
        "sort_type": "综合排序", "publish_time": "不限", "cursor0": "0",
        # 对齐 ContentFindAgentNew CONTENTFIND_DOUYIN_DEFAULT_ACCOUNT_ID；缺它抖音搜索 code 10000
        "account_id": "771431222",
    },
}


class SearchError(RuntimeError):
    pass


def parse_search_response(response: Any, *, platform: str = "xiaohongshu") -> tuple[list[str], bool, str]:
    """信封 {code,msg,data:{has_more,next_cursor,data:[...]}} → (content_ids, has_more, next_cursor)。
    条目 id 字段按平台取（小红书 id / 抖音 aweme_id）。code!=0 抛 SearchError。"""
    if not isinstance(response, dict):
        raise SearchError("bad_response: not a dict")
    code = response.get("code")
    if code not in (0, "0"):
        raise SearchError(f"business_error: code={code} msg={response.get('msg')}")
    data = response.get("data") or {}
    items = data.get("data") or []
    id_key = (PLATFORM_SEARCH.get(platform) or {}).get("id_key", "id")
    ids = [it.get(id_key) for it in items if isinstance(it, dict) and it.get(id_key)]
    return ids, bool(data.get("has_more")), str(data.get("next_cursor") or "")


def search_keyword(
    keyword: str,
    *,
    platform: str = "xiaohongshu",
    content_type: str = "图文",
    sort_type: Optional[str] = None,       # None → 平台默认（小红书:综合 / 抖音:综合排序）
    publish_time: Optional[str] = None,     # None → 平台默认（抖音:不限）
    account_id: Optional[str] = None,       # None → 平台默认（抖音:771431222；小红书无）
    limit: int = 5,
    settings: Optional[Settings] = None,
    http_client: Any = None,
    rate_limiter: Optional[RateLimiter] = None,
    env_file: str = ".env",
    max_pages: int = 10,
) -> list[str]:
    """搜索 → content_id 列表（翻页直到够 limit 或 has_more=False；按 id 去重、截断 limit）。
    平台专属参数（path/sort_type/publish_time/起始cursor/account_id）取自 PLATFORM_SEARCH，
    显式入参可覆盖。复用 crawler.RateLimiter（'{platform}_keyword' bucket）。失败抛 SearchError。"""
    cfg = PLATFORM_SEARCH.get(platform)
    if not cfg:
        raise SearchError(f"unsupported platform: {platform}")
    settings = settings or Settings.from_env(env_file)
    sort_type = cfg["sort_type"] if sort_type is None else sort_type
    publish_time = cfg["publish_time"] if publish_time is None else publish_time
    account_id = cfg["account_id"] if account_id is None else account_id
    rate_limiter = rate_limiter or RateLimiter()

    owns_client = http_client is None
    client = http_client or httpx.Client()
    out: list[str] = []
    seen: set[str] = set()
    cursor = cfg["cursor0"]
    try:
        for _ in range(max_pages):
            rate_limiter.wait(f"{platform}_keyword")
            url = urljoin(settings.crawler_base_url, cfg["path"])
            body = {"keyword": keyword, "content_type": content_type,
                    "sort_type": sort_type, "publish_time": publish_time, "cursor": cursor}
            if account_id:
                body["account_id"] = account_id   # 抖音必带；小红书不带
            try:
                resp = client.post(url, json=body,
                                   headers={"Content-Type": "application/json"},
                                   timeout=settings.crawler_timeout)
                resp.raise_for_status()
                data = resp.json()
            except httpx.HTTPError as exc:
                raise SearchError(f"http_error: {exc}") from exc
            except ValueError as exc:
                raise SearchError("bad_json") from exc
            ids, has_more, next_cursor = parse_search_response(data, platform=platform)
            for cid in ids:
                if cid not in seen:
                    seen.add(cid)
                    out.append(cid)
                    if len(out) >= limit:
                        return out
            if not has_more or not next_cursor or next_cursor == cursor:
                break
            cursor = next_cursor
    finally:
        if owns_client:
            client.close()
    return out[:limit]


# ---------- 微信公众号搜索（回的是带 url/封面的文章，不是纯 id，故单列）----------
_EM = re.compile(r"<[^>]+>")


def parse_weixin(response: Any, *, limit: int = 5) -> list[dict]:
    """微信搜索回包 {code,data:{data:[{title,url,cover_url,nick_name,time}]}} → 文章列表（title 去 <em> 标记）。"""
    if not isinstance(response, dict):
        raise SearchError("bad_response: not a dict")
    if response.get("code") not in (0, "0"):
        raise SearchError(f"business_error: code={response.get('code')} msg={response.get('msg')}")
    items = ((response.get("data") or {}).get("data")) or []
    out: list[dict] = []
    for it in items:
        if not isinstance(it, dict) or not it.get("url"):
            continue
        out.append({"title": _EM.sub("", it.get("title", "")).strip(),
                    "url": it.get("url", ""), "cover_url": it.get("cover_url") or "",
                    "nick_name": it.get("nick_name") or "", "time": it.get("time") or ""})
        if len(out) >= limit:
            break
    return out


def search_weixin(
    keyword: str,
    *,
    limit: int = 5,
    settings: Optional[Settings] = None,
    http_client: Any = None,
    rate_limiter: Optional[RateLimiter] = None,
    env_file: str = ".env",
) -> list[dict]:
    """微信公众号搜索：query → 文章列表（带 url/封面，无需 detail）。失败抛 SearchError。"""
    settings = settings or Settings.from_env(env_file)
    rate_limiter = rate_limiter or RateLimiter()
    owns_client = http_client is None
    client = http_client or httpx.Client()
    try:
        rate_limiter.wait("weixin_keyword")
        url = urljoin(settings.crawler_base_url, "/crawler/wei_xin/keyword")
        try:
            resp = client.post(url, json={"keyword": keyword, "cursor": "0"},
                               headers={"Content-Type": "application/json"},
                               timeout=settings.crawler_timeout)
            resp.raise_for_status()
            data = resp.json()
        except httpx.HTTPError as exc:
            raise SearchError(f"http_error: {exc}") from exc
        except ValueError as exc:
            raise SearchError("bad_json") from exc
        return parse_weixin(data, limit=limit)
    finally:
        if owns_client:
            client.close()